Bytedance Seedream V4.5 Edit Sequential API Documentation

Bytedance Seedream V4.5 Edit Sequential API Documentation

Playground

Seedream 4.5 Edit Sequential performs multi-image editing while locking character and object identity across shots. It detects main subjects, preserves continuity, and applies controlled edits with up to 4K output. Ready-to-use REST inference API, best performance, no cold starts, affordable pricing.

Features

Seedream 4.5 Edit Sequential is ’s multi-image editing model designed to apply the same edit across a whole set of images. It automatically tracks the main subject through the series, keeps identity stable, and generates clean, high-resolution results—ideal for campaigns, product sets, and character line-ups.

Model highlights

• Multi-image subject tracking – Detects the main subject across all input images and treats them as the same person or object.
• Character consistency lock – Preserves facial structure, proportions, and overall identity across every edited output.
• High reference fidelity – Maintains lighting, colour balance, and key visual details while applying the requested changes.
• Controlled, repeatable edits – One prompt drives a consistent transformation across the entire batch.
• 4K-ready resolution – Supports sizes up to 8192 × 8192 for print-adjacent and hero visual use.
• Production quality – Sharp edges, low artifacts, and stable composition suitable for professional workflows.

Best suited for

• Batch portrait editing with a fixed style or look
• Product series images that must feel like one coherent set
• Brand or ad campaign iterations with the same model or hero product
• Character design variations (outfits, moods, lighting)
• E-commerce catalog refreshes and seasonal updates
• Social / marketing visual series where continuity matters

Pricing

Billing is per output image, scaled by the max_images you request.

• Base price: $0.04 per image
• Formula: total_price = $0.04 × max_images

Example costs:

How to use

1. Upload source images Add the images you want to edit sequentially (all should contain the same main subject or product).

2. Write the edit prompt Describe the shared change you want across the whole set, e.g. “Change outfit to a black suit, add soft studio lighting, keep poses and background the same.”

3. Set max_images Specify how many edited outputs you want the model to generate from your input set.

4. Choose size Select the target resolution, up to 8192 × 8192 for maximum detail.

5. Run and review Submit the job, inspect the edited series, and optionally refine the prompt for another pass.

Tips for best results

• Use clear, global instructions in the prompt (“add winter outfit and snow ambience”) rather than per-image directions.
• Keep input images reasonably consistent in framing and lighting so the model can lock onto the same subject.
• Put your cleanest, clearest reference image first; the model tends to rely on it most strongly for identity.
• For campaign work, generate at the highest resolution you need once, then downscale for web or social formats.

Note

• Please set the max_image first, and then input how many images you want to generate in prompt! Such as:

• max_image = 4. Prompt: I want to generate 4 images… + (your prompt)

More Models to Try!

• Nano Banana Pro

Google’s ultra, fast text-to-image model for generating many ideas from scratch; great for large batches of new images, not for editing existing photo series.

• Seedream V4

’s high-resolution text-to-image generator with rich detail and diverse styles; ideal when you want new scenes in the Seedream aesthetic rather than editing your own photos.

• Qwen Image Edit Plus

Single-image, prompt-based editing with strong semantic understanding; perfect for one-off or small-batch edits where you don’t need strict identity matching across many images.

Authentication

For authentication details, please refer to the Authentication Guide.

API Endpoints

Submit Task & Query Result

set -euo pipefail

export WAVESPEED_API_KEY="your-api-key"

REQUEST_BODY=$(cat <<'JSON'
{
  "prompt": "A cinematic ocean wave at sunrise, highly detailed",
  "images": [
    "https://interactive-examples.mdn.mozilla.net/media/cc0-images/painted-hand-298-332.jpg"
  ],
  "max_images": 1
}
JSON
)

# 1. Submit the prediction once. Never retry this request automatically:
# a retry after an unclear failure can create a second, billed task.
CHECK='The task may already exist. Check History or the predictions list before submitting again.'
if ! SUBMIT_RESPONSE=$(curl --silent --show-error --fail-with-body \
  --connect-timeout 10 --max-time 60 \
  -X POST "https://api.wavespeed.ai/api/v3/bytedance/seedream-v4.5/edit-sequential" \
  -H "Authorization: Bearer ${WAVESPEED_API_KEY}" \
  -H "Content-Type: application/json" \
  -d "${REQUEST_BODY}"); then
  printf 'Submit failed. %s\n' "${CHECK}" >&2
  exit 1
fi

TASK=$(printf '%s' "${SUBMIT_RESPONSE}" | jq 'if type == "object" and has("data") then .data else . end')
PREDICTION_ID=$(printf '%s' "${TASK}" | jq -r '.id // empty')
if [ -z "${PREDICTION_ID}" ]; then
  printf 'Submission response did not contain a prediction id. %s\n' "${CHECK}" >&2
  exit 1
fi
RESULT_URL="https://api.wavespeed.ai/api/v3/predictions/${PREDICTION_ID}/result"

# 2. Poll until the prediction reaches a final status.
BODY_FILE=$(mktemp)
trap 'rm -f "${BODY_FILE}"' EXIT
while true; do
  # Transient errors (429, 5xx) are retried, honoring Retry-After.
  HTTP_CODE=$(curl --silent --show-error --fail-with-body \
    --connect-timeout 10 --max-time 60 --retry 5 --retry-delay 2 \
    --output "${BODY_FILE}" --write-out '%{http_code}' \
    "${RESULT_URL}" \
    -H "Authorization: Bearer ${WAVESPEED_API_KEY}" || true)
  case "${HTTP_CODE}" in
    2??) ;;
    000|429|5??)  # dropped connection or temporary error: keep polling
      printf 'Task %s: HTTP %s, polling again in 10s\n' "${PREDICTION_ID}" "${HTTP_CODE}" >&2
      sleep 10; continue ;;
    *) printf 'Task %s: HTTP %s\n' "${PREDICTION_ID}" "${HTTP_CODE}" >&2
      cat "${BODY_FILE}" >&2; exit 1 ;;
  esac
  RESULT=$(jq 'if type == "object" and has("data") then .data else . end' "${BODY_FILE}")
  STATUS=$(printf '%s' "${RESULT}" | jq -r '.status // empty')

  case "${STATUS}" in
    completed) printf '%s\n' "${RESULT}" | jq '.outputs'; break ;;
    failed|cancelled|timeout|deleted)
      printf 'Task %s ended with status %s\n' "${PREDICTION_ID}" "${STATUS}" >&2
      printf '%s\n' "${RESULT}" | jq . >&2; exit 1 ;;
    *) sleep 2 ;;
  esac
done

Parameters

Task Submission Parameters

Request Parameters

Every model also accepts the global request options enable_sync_mode and enable_base64_output; see their pages for details.

Response Parameters

Result Request Parameters

Result Response Parameters